Abstract :
In this paper the authors investigate the authorship of several short historical texts that are written by ten ancient Arabic travelers: this Arabic dataset, which was collected by the authors in 2011, is called AAAT dataset. Several experiments of authorship attribution are conducted on these Arabic texts, by using different lexical features such as words, word-big rams, word-trig rams, word-tetra grams and rare words. Furthermore, seven different classifiers are employed, namely: Manhattan distance, Cosine distance, Stamatatos distance, Camberra distance, Multi Layer Perceptron (MLP), Sequential Minimal Optimization based Support Vector Machine (SMO-SVM) and Linear Regression. For the evaluation task, several experiments of authorship attribution are conducted on the AAAT dataset by using the different quoted features and classifiers. Results show good attribution performances with an optimal score of 80% of good authorship attribution. Moreover, this investigation has revealed interesting results concerning the Arabic language and more particularly for the short texts.
Keywords :
multilayer perceptrons; natural language processing; optimisation; pattern classification; regression analysis; support vector machines; text analysis; AAAT dataset; Arabic dataset; Arabic language; Camberra distance; Cosine distance; MLP; Manhattan distance; SMO-SVM; Stamatatos distance; authorship attribution of ancient Arabic text; classifiers; lexical features; linear regression; multilayer perceptron; rare words; sequential minimal optimization based support vector machine; short historical Arabic text; word-big rams; word-tetra grams; word-trig rams; Educational institutions; Encyclopedias; Linear regression; Support vector machine classification; Training; Vectors; Arabic language; Artificial Intelligence; Authorship Attribution; Computational Linguistics; Lexical Features; Rare words; Text-mining; Word N-Grams;
